Study On The Prediction Method Of Postoperative Recurrence Of Hepatocellular Carcinoma Based On Texture Analysis Of CT Liver Images

Hepatocellular carcinoma(HCC)is very harmful to human body.At present,the main method of HCC treatment is partial hepatectomy,which is easy to recurrence,leading to an unsatisfactory outcome.Computed tomography(CT)imaging is a kind of detecting method for HCC,using texture analysis to get quantifiable parameters from the regions of interest(ROI)of CT images.The parameters produce imaging biomarker for prognosis,providing support for recurrence detection and diagnosis after treatment.Based on liver texture analysis of CT images,the prediction method of recurrence of HCC patients within two years was studied.Three parts are included in the method: feature extraction,feature selection and classifier design.For feature extraction,features are extracted from original CT images and the corresponding of four images processed by different scales of Laplacian of Gaussian filter.A local contrast enhancement will be done on the images before extraction.Texture feature extraction from 2D and 3D regions of interest,which include image intensity,image gradient,Gabor,local binary pattern histogram Fourier and gray-level co-occurrence matrix.For feature selection,an embedded feature selecting method is applied to collect frequency of features using multiple sampling.Features with frequency larger than pre-defined threshold will be selected.For classifier design,the support vector machine classification model will be adopted.By importing selected features to the support vector machine,estimates of recurrence risk after the treatment will be given.Using the above method to extract and select features from multiphase original CT images and images processed by Laplacian of Gaussian filter and producing recurrence estimates by support vector machine with selected features.The results indicate that after selecting features extracted from both original CT images and images processed by Laplacian of Gaussian filter,accuracy of estimating recurrence risk after HCC treatment becomes relatively ideal.